Increasing deterministic behavior of mobile robots by adding a safety layer

Mark Bruijn
Presentation MSc presentation
Date 2019-02-25
Time 11:00
Location Hal B 2A

Tele-operated mobile robots have a potentially high value for emergency services. However, for most mobile robots, onboard failures will currently result in random behavior of the mobile robot. This random behavior does not match the operator's expected response, which implies non-deterministic behavior. Operators require deterministic behavior of mobile robots at all times, even after onboard failures. If this requirement can be met, overall support for using mobile robots will increase, fewer emergency operations will fail and dangerous consequences are prevented.

In this thesis, I will research how to increase deterministic behavior of mobile robots by implementing a safety layer. Inspired by critical chemical processes, the safety layer mitigates consequences. This gives time to the safety layer to eliminate random behavior. The advantage of this approach is that any unknown failure cause is processed by the safety layer. To make the safety layer fit for implementation on any mobile robot, the safety layer is generic. The safety layer uses a watchdog onboard the mobile robot to detect onboard failures. The independent part of the safety layer is responsible for providing a response and eliminating random behavior. During onboard failures, the safety layer takes over the robot's controls. The safety layer has sensors to effectuate its responses. After testing the functionalities, the effect of the safety layer on a mobile robot must be determined.

An increase in deterministic behavior can be realized using a safety layer. The safety layer successfully mitigates the consequences of onboard failures. This includes onboard-computer failures, to which protection is in general currently not implemented. The safety layer is successfully tested for 25 hours to ensure it does not affect normal operations of a mobile robot. A careful estimation indicates that the safety layer reduces the probability of consequences with 2% to 7% for the mobile robot at the University of Twente.

Posted on Monday, February 4, 2019